THE AUK A QUARTERLY JOURNAL OF ORNITHOLOGY VOL. 105 JULY 1988 No. 3 A CLASSIFICATION OF THE LIVING BIRDS OF THE WORLD BASED ON DNA-DNA HYBRIDIZATION STUDIES CHARLESG. SIBLEY,• JON E. AHLQUIST,2 AND BURTL. MONROE JR.3 •TiburonCenter for EnvironmentalStudies, San Francisco State University, Box855, Tiburon,California 94920 USA, 2Departmentof Zoologicaland Biomedical Sciences, Ohio University, Athens, Ohio 45701 USA, and 3Departmentof Biology,University of Louisville,Louisville, Kentucky 40292 USA ABSTRACT.--Wepresent'a Classification of the living birdsof the world basedon the results of DNA-DNA hybridizationstudies. Several family-group names are presentedformally for the first time; someof these nameshave appearedearlier in various publications.Received 21 January1988, accepted30 March 1988. THIS classification, and the phylogeny on The dendrograms were constructedby the which it is based,were first presentedas posters method of average linkage, or UPGMA (un- at the XIX International Ornithological Con- weighted pair group method using arithmetic gressin Ottawa, Canada,in June 1986,and later averages,Sheath and Soka11973:230). The DNAs at the A.O.U. annual meetings in 1986 (Stark- of 1,058 species,of which 310 were radio-la- ville, Mississippi)and 1987(San Francisco,Cal- beled tracers,were used to reconstructthe phy- ifornia). Figures 1-5 depict the branching pat- logeny. Groupsat and abovethe ordinal level tern of the phylogeny above the level of were clustered from several overlapping ma- subfamilies.Dendrograms of the completeDNA- trices. For example, in the Passeriformescom- based phylogeny will be included in a forth- plete matriceswere preparedfor the suboscines, comingbook (Sibley and Ahlquist in prep.). We the Corvida and the Passerida,and among these present the classificationat this time because matricesenough measurementswere made to parts of it have been, or will be, used in works establishthe positionsof these groups relative by other authors. to one another and to the non-passerines.The We believe that DNA-DNA hybridization major groups were clustered on the basis of comparisonsprovide data that may be used to complete, or nearly complete, matricesconsist- reconstructthe branching pattern of the phy- ing of the radio-labeled species.At least one logeny of living birds. The delta T50H values speciesin each group in Figs. 1-5 was radio- (e.g. Sibley and Ahlquist 1981, 1983, 1986) are labeled and reciprocals between nearly all believed to be relative to time, but we no longer branches were made. The averaging effect of assumethat there is a simple relationship be- UPGMA smooths out some of the effects of dif- tween delta values and absolute time. Instead, ferent averagerates of genomic evolution, but it seemslikely that the averagerate of nucleo- relative rate tests were used to determine the tide substitutionmay be correlatedwith factors branching order of taxa. related to reproduction;for example,age at first Hennig (1966:160) proposed the age of origin breeding, generation time, number of cell di- as the ideal basisfor the ranking of taxonomic visions in the germline, and possibly other fac- categories.The application of this principle tors. This subjectis discussedby Ahlquist et al. would producea classificationin which differ- (1987), Catzeflis et al. (1987), Sibley and Ahl- ent lineageswith the sameage of origin would quist (1987d), and Sibley et al. (1987). be placed in the same categoricalrank, thus 409 The Auk 105: 409-423. July 1988 410 $IBLEY,AHLQUlST, AND MONROE JR. [Auk,Vol. 105 ORDER SUPERORDER PARVCLASS INFRACLASS STinamir............... ormes 1Ratitae •> Galliformes Galloansera• m -•25.9 (992•221.6 Crac{forrnesAnserlforrnes ]-- GallomorphaeAnserimorphae I O ................................... Turniclformes [.•24A[63)• ........... ] coraciae 25;[256• Cu;luf;r;Ss. Cuculimor haeColiae Trochiliformes• Apod•morphae • •)(2n .....i...... • g p Passerae Fig. 1. Major divisions of the classAves. The numbers, e.g. 28.0 (316), indicate that the average delta T5oHvalue of 28.0 is basedon the delta Ts0Hvalues of 316 DNA-DNA hybridsbetween taxa on the two sides of the branch node. achievingthe desirablegoal of categoricalequiv- categoricalequivalence into our classification alence.Hennig (1966: 161-182) discussedthe by assigningranks on the basisof delta T50H problemsassociated with determiningthe ages values (Table 1). The delta values express de- of origin of systematicgroups and concluded greesof genomicdivergence and, sincethe av- that the availablemethods were inadequate.The eragerate of divergenceis relative to time, this presentmethods still fall shortof the ideal, but procedureprovides an approachto categorical we have tried to incorporate the principle of equivalencebased on time of origin. The fact I I • THEFAMILIES OFNON-PASSERINE BIRDS Fig. 2. Families of non-Passerine birds, Struthionidae to Dacelonidae. July1988] ClassificationofLiving Birds 411 • 24.5(132) 19.1 (76} Eurypygidae 17.5 (27} • 14.5(6} CrotophagidaeNeornorphidae Otididae • 16.4(5) Opisthocomidae 16 9 (47} Cariarnidae • 176 (53) Coccyzidae Rhynochetidae ß P 15.0 (2} Psophiidae 14.• Centropodidae He]iornithidae Gruidae Psittacidae Pteroclidae • . •95 (4} Hem procnidae Jacanidae 12.5 (5} Rostratulidae 14.1 (49) Thinocorldae -- 10.8 (8} onomidae (46} )acidae • 13.6(7) Tytonidae Gharadriidae lO.6 (92} Bur hinidae 19.1{55) Aegothelidae 11 5 (27) Chionldidae 18.8 (26} Podargidae 10.3 {2} Batrachostornidae (87) Glareolidae • Nyctibiidae (34} Laridae Falconidae • 169{10) 12.3 (3} Caprimulgidae Sagittariidae Ipitridae __ Columbidae Podicipedidae Phaethontidae (112) Sulidae 11.0(5} gidae 12.1(30) Phalacrocoracidae (323} Ardeldae Scopidae Phoenicopterldae Fig. 3. Families of non-passerinebirds, Coliidae Threskiornithidae to Columbidae. (104} Ciconiidae 10.1(82) Pelecanidae 10.9 (169} Fregatidae 10.7(50} Spheniscidae that different lineagesevolve at somewhatdif- Gaviidae 10.0 (47} Procellariidae ferent averagerates introduces an error of un- certain magnitude. 21.6(129) Passeriformes The Turnicidaemay presenta specialprob- Fig. 4. Familiesof non-passerinebirds, Rallidae lem. The 27 delta T50H values between Turnix to Procellariidae, and the position of the Passeri- and other taxa are large, ranging from 22.7 to formes. the pectoralsandpiper (Calidris melanoto•) to 29.8 to the ostrich (Struthiocamelus); average = 27.0. Does the Turnix problem cast doubt on the The rangeof 7.1 is unusuallylarge. The smaller positionsof all taxa in the tree? We think not, delta values are between Turnix and members becausenearly all other groupsconsist of many of the Gruiformes and Charadriides;the larger speciesand severallevels of branchings,but the delta values are between Turnix and the ratites, memberscluster in spite of somewhatdifferent tinamous, galliforms and anseriforms. Mem- agesat maturity. For example,members of the bers of severalother non-passerinegroups pro- Procellariidaerange in ageat first breedingfrom duced delta values between the extremes. Thus, ca. 3-4 years (Storm-Petrels)to more than !0 Turnixseems to have no close living relatives, years(large albatrosses).Differences in branch but Turnixspecies begin to breed at the age of lengths are revealedby relative rate tests,but three to five months (Bruning 1985), therefore all procellariids are closer to one another than it is possiblethat their averagerate of genomic to any outsidegroup. Similarly, the estrildine evolution is faster than the rate(s)among their finchesbegin to breed when lessthan one year nearestrelatives. If so, the branch length we of age,but they clusterwith the other members observemay be longer by an unknown factor of the Passeridae and are closest to the yiduines, and the correctposition of the branch node may as expected. be elsewhere in the tree, possibly among the Thus, it is possible that the Turnicidae are gruiformswhere the Turnicidaehave often been actually as isolatedas their position in Figs. ! assigned.The problemis not insoluble,but will and 2 suggests,but their exceptionallyearly require additional comparisonsand, perhaps, maturity, and possiblerapid DNA evolution, the developmentof a correctionfor different raise a question that must be considered, but agesat first breeding. For now we leave the cannot be resolved at this time. Turnicidae in the position indicated in Figs. ! Hennig (!966: !54-!60) proposedthat the sis- and 2, but designatethe group as incertaesedis. ter branchesat each dichotomy should receive 412 SmLE¾,AHLQUIST, AND MONROE JR. [Auk,Vol. 105 Suborder Suborder Tyranni Passeri Parvorder Parvorder Corvida Passerida •..=__='E •o.- ß THE FAMILIES OF PASSERINE BIRDS t9.7 [89) Fig. 5. Familiesof passefinebirds. the samecategorical rank, and that each differ- When the classification was first developed ent level of branching should be assigneda the categorieswere basedon units of 10 million different rank, arranged accordingto their rel- yearsand equatedwith delta T50Hvalues on the ative age of origin. From Figs. 1-5 it is apparent basisof delta T•0H 1.0 = 4.5 million years (e.g. that the DNA-based phylogeny contains nu- Sibleyand Ahlquist 1983, 1985a, b). It now seems merous dichotomies at many levels and that it likely that this calibration factor is not appli- would be impractical to assign a different cat- cable to all birds, but only to those with ages egoricalname to every branchnode in the phy- at first breeding of ca. 2-4 years. Speciesthat logeny. Therefore,to accommodatethe several branchesthat often occurwithin a single rank TABLE1. Categories,endings